Keras Prediction after test values - python

I am currently trying to build neuronal network to be able to predict time series, but the question is, is it possible to predict further than just the test dataset. I mean, for my example, I have a dataset of about 3000 values, from which I keep 90% for training and 10% for testing. Then When I compare the prediction with the actual test value, it corresponds, but is it possible for instance to ask the program to predict the next 500 values (i.e. from 3001 to 3500) ?
Here is a snipper of the code I use.
import csv
import numpy as np
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import matplotlib.pyplot as plt
from keras.layers.core import Dense, Activation, Dropout
from keras.layers.recurrent import LSTM, GRU
from keras.models import Sequential
from keras import optimizers
from sklearn.svm import SVR
from sklearn.model_selection import GridSearchCV
from sklearn.model_selection import learning_curve
from sklearn.kernel_ridge import KernelRidge
import time
from sklearn.preprocessing import MinMaxScaler
sc = MinMaxScaler(feature_range = (-1, 1))
def load_data(datasetname, column, seq_len, normalise_window):
# A support function to help prepare datasets for an RNN/LSTM/GRU
data = datasetname.loc[:,column]
sequence_length = seq_len + 1
result = []
for index in range(len(data) - sequence_length):
result.append(data[index: index + sequence_length])
result = np.array(result)
result.reshape(-1,1)
training_set_scaled = sc.fit_transform(result)
print (result)
#Last 10% is used for validation test, first 90% for training
row = round(0.9 * training_set_scaled.shape[0])
train = training_set_scaled[:int(row), :]
#np.random.shuffle(train)
x_train = train[:, :-1]
y_train = train[:, -1]
X_test = training_set_scaled[int(row):, :-1]
y_test = training_set_scaled[int(row):, -1]
print ("shape train", x_train)
print ("shape train", X_test)
x_train = np.reshape(x_train, (x_train.shape[0], x_train.shape[1], 1))
X_test = np.reshape(X_test, (X_test.shape[0], X_test.shape[1], 1))
return [x_train, X_test, y_train, y_test]
def build_model():
model = Sequential()
layers = {'input': 100, 'hidden1': 150, 'hidden2': 256, 'hidden3': 100, 'output': 10}
model.add(LSTM(
50,
return_sequences=True,
input_shape=(200,1)
))
model.add(Dropout(0.2))
model.add(LSTM(
layers['hidden2'],
return_sequences=True,
))
model.add(Dropout(0.2))
model.add(LSTM(
layers['hidden3'],
return_sequences=False,
))
model.add(Dropout(0.2))
model.add(Activation("linear"))
model.add(Dense(
output_dim=layers['output']))
start = time.time()
model.compile(loss="mean_squared_error", optimizer="adam")
print ("Compilation Time : ", time.time() - start)
return model
dataset = pd.read_csv(
'data.csv')
X_train, X_test, y_train, y_test = load_data(dataset, 'mean anomaly', 200, False)
model = build_model()
print ("train",X_train)
print ("test",X_test)
model.fit(X_train, y_train, batch_size=256, epochs=1, validation_split=0.05)
predictions = model.predict(X_test)
predictions = np.reshape(predictions, (predictions.size,))
plt.figure(1)
plt.subplot(311)
plt.title("Actual Test Signal w/Anomalies & noise")
plt.plot(y_test)
plt.subplot(312)
plt.title("predicted signal")
plt.plot(predictions, 'g')
plt.subplot(313)
plt.title("training signal")
plt.plot(y_train, 'b')
plt.plot(y_test, 'y')
plt.legend(['train', 'test'])
plt.show()
I have read that I should increase the output dim of the dense layer to get more than 1 predicted value, or increase the size of my window in the load data function ?
Here is the result, the yellow plot is supposed to be after the blue one, it respresents my input test data, the first plot is a zoom on this data and the second one the prediction.

If you want to predict the output value of your serie at t+x based on data at time t, the data you need to feed to the network should already have this format.
Time series data formating :
If you have 3000 data point and want to predict the output value for the next "virtual" 500 point you should offset the output value by this amount. For exemple :
In your dataset, your 500th data point correspond to the 500th output value. If you want to predict "future" values then the 500th data point should have the 1000th output value. You can do this in pandas with the shift function. Be aware that you will loose the last 500 data point by doing so, has they will no longer have an output value.
Then when you predict on data point xi you'll have the output value yi+500. You should find some basic guides for time serie forecasting on sites like machinelearningmastery
Good pratice for model evaluation :
If you want to better evaluate the quality of your model, first find some metrics that suits your problem and try to increase test set percenatage. While graphics are a good way to visualise result, they can be deceiving, try combining them with some metrics ! (be carefull with Mean Squarred Error, it can give you a biased score with value in the range [-1;1] as the square of an error in this range will always be less than the acutal error, try Mean Absolute Error instead)
Data leakage when scalling data :
While scalling data is usually a good thing you need to be carefull doing so. You comited something called a data leak. You used scalling on the whole data set before splitting into training and test set. Further reading about this data leak.
Update
I think i misunderstood your problem.
If you want to "predict further than just the test dataset" you will need some unseen/new data to make more prediction. The test set is only made to evaluate the performance of the learning phase.
Now if you want to predict further than just the next step (this won't allow you to "predict further than just the test dataset" because of the way you change your dataset, see bellow) :
Your model as it's made will only ever predict the next step.
In your example you feed to the algorithm series of lenght 'seq_len' and give them as output the value right after the end of those series. If you want your algorithm to learn to predict in more than one step into the future you y_train must have value at the corresponding time in the future, example :
x = [0,1,2,3,4,5,6,7,8,9,10,...]
seq_len = 5
step_to_predict = 5
So to predict not one step into the future but five, your series will have to look like this :
x_serie_1 = [0,1,2,3,4]
y_serie_1 = [9]
x_serie_2 = [1,2,3,4,5]
y_serie_2 = [10]
This is a way to get your model to learn how to make predictions further into the future than just the next step.

Related

Getting a probability distribution curve from a TensorFlow model

I am trying to learn working with TensorFlow and so I was trying to make a probablistic ML model to get the probability distribution of the next day stock price based on the last n days price sequence, and when doing so I managed to predict the next day's price but not getting a probability distribution of the model. How do I get the curve which the model predictions are based on from the TensorFlow model?
This is the code i have got so far that predicts the actual price for the next day (using this video: https://www.youtube.com/watch?v=PuZY9q-aKLw):
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from datetime import datetime
from datetime import timedelta
from sklearn.preprocessing import MinMaxScaler
import tensorflow as tf
import tensorflow_probability as tfp
tfd = tfp.distributions
tfpl = tfp.layers
tfk = tf.keras
tfkl = tf.keras.layers
# preparing data
min_max_scaler = MinMaxScaler(feature_range=(0,1)) # a scaler object which normalizes the number between 0 and 1 in relation to the rest of the dataset
close_prices = price_data.loc[:,'Close'].values # all the close prices of the price_data df
scaled_close_prices = min_max_scaler.fit_transform(close_prices.reshape(-1,1)) # all the close prices, normalized between 1 and 0
n = 60 # the number of days that are used to determine the next value
# x contains the last n days before the predicted day which is y
X_train = [] # 70% of the dataset
y_train = []
X_test = [] # 30% of the dataset
y_test = []
for x in range(n, int(len(scaled_close_prices)*0.7)):
X_train.append(scaled_close_prices[x-n:x,0])
y_train.append(scaled_close_prices[x,0])
for x in range(int(len(scaled_close_prices)*0.7), len(scaled_close_prices)):
X_test.append(scaled_close_prices[x-n:x,0])
y_test.append(scaled_close_prices[x,0])
X_train, y_train = np.array(X_train), np.array(y_train)
X_test, y_test = np.array(X_test), np.array(y_test)
X_train = np.reshape(X_train, (X_train.shape[0], X_train.shape[1], 1))
X_test = np.reshape(X_test, (X_test.shape[0], X_test.shape[1], 1))
# building the model
model = tfk.Sequential()
model.add(tfkl.LSTM(units=50, return_sequences=True, input_shape=(X_train.shape[1], 1)))
model.add(tfkl.Dropout(0.2))
model.add(tfkl.LSTM(units=50, return_sequences=True))
model.add(tfkl.Dropout(0.2))
model.add(tfkl.LSTM(units=50))
model.add(tfkl.Dropout(0.2))
model.add(tfkl.Dense(units=1))
# compiling model
model.compile(optimizer=tfk.optimizers.Adam(learning_rate=0.05),
loss='mean_squared_error',
metrics=[])
# fitting model
model.fit(X_train, y_train, epochs=25, batch_size=32)
# testing model
y_predicted = model.predict(X_test)
y_predicted = min_max_scaler.inverse_transform(y_predicted).reshape(-1)
Probabilistic modelling one of the things that tensorflow_probability (tfp) can do.
It can do probabilistic time series forecasting - e.g. see https://www.tensorflow.org/probability/examples/Structural_Time_Series_Modeling_Case_Studies_Atmospheric_CO2_and_Electricity_Demand
... but I don't believe it includes a variational LSTM in its toolbox, and they may be close to the current research frontier. e.g. see https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7842932/
As written, this model is not especially probabilistic, per se, but it can be viewed as outputting an estimate of the mean of a gaussian distribution with unit variance; this is implicit in using a squared error loss. To make this more concrete, you could instantiate a tfd.Normal distribution with the output of your model as the loc parameter and scale=1., and call prob(...) to evaluate the pdf of this distribution. I'm not sure if this is what you're looking for, though.

Not able to get correct results even though mean absolute error is low

import pandas as pd
df=pd.read_csv('final sheet for project.csv')
features=['moisture','volatile matter','fixed carbon','calorific value','carbon %','oxygen%']
train_data=df[features]
target_data=df.pop('Activation energy')
X_train, X_test, y_train, y_test = train_test_split(train_data,target_data, test_size=0.09375, random_state=1)
standard_X_train=pd.DataFrame(StandardScaler().fit_transform(X_train))
standard_X_test=pd.DataFrame(StandardScaler().fit_transform(X_test))
y_train=y_train.values
y_train = y_train.reshape((-1, 1))
scaler = MinMaxScaler(feature_range=(0, 1))
scaler = scaler.fit(y_train)
normalized_y_train = scaler.transform(y_train)
y_test=y_test.values
y_test = y_test.reshape((-1, 1))
scaler = MinMaxScaler(feature_range=(0, 1))
scaler = scaler.fit(y_test)
normalized_y_test = scaler.transform(y_test)
model=keras.Sequential([layers.Dense(units=20,input_shape=[6,]),layers.Dense(units=1,activation='tanh')])
model.compile(
optimizer='adam',
loss='mae',
)
history = model.fit(standard_X_train,normalized_y_train, validation_data=(standard_X_test,normalized_y_test),epochs=200)
I wish to create a model to predict activation energy using some features . I am getting training loss: 0.0629 and val_loss: 0.4213.
But when I try to predict the activation energies of some other unseen data ,I get bizarre results. I am a beginner in ML.
Can someone please help what changes can be made in the code. ( I want to make a model with one hidden layer of 20 units that has activation function tanh.)
You should not use fit_transform for test data. You should use fit_transform for training data and apply just transform to test data, in order to use the same parameters for training data, on the test data.
So, the transformation part of your code should change like this:
scaler_x = StandardScaler()
standard_X_train = pd.DataFrame(scaler_x.fit_transform(X_train))
standard_X_test = pd.DataFrame(scaler_x.transform(X_test))
y_train=y_train.values
y_train = y_train.reshape((-1, 1))
y_test=y_test.values
y_test = y_test.reshape((-1, 1))
scaler_y = MinMaxScaler(feature_range=(0, 1))
normalized_y_train = scaler_y.fit_transform(y_train)
normalized_y_test = scaler_y.transform(y_test)
Furthermore, since you are scaling your data, you should do the same thing for any prediction. So, your prediction line should be something like:
preds = scaler_y.inverse_transform(
model.predict(scaler_x.transform(pred_input)) #if it is standard_X_test you don't need to transform again, since you already did it.
)
Additionally, since you are transforming your labels in range 0 and 1, you may need to change your last layer activation function to sigmoid instead of tanh, or even may better to use an activation function like relu in your first layer if you are still getting poor results after above modifications.
model=keras.Sequential([
layers.Dense(units=20,input_shape=[6,],activation='relu'),
layers.Dense(units=1,activation='sigmoid')
])

Solar power prediction using Keras

Dataset:
The PV Yield (kWh) is my output. My model is suppose to predict this.
This is what I have done. I have attached the image of the dataset. From AirTemp to Zenith is my X and Y is PV Yield(KW/H).
df=pd.read_csv("Data1.csv")
X=df.drop(['Date-PrimaryKey','output-PV Yield (kWh)'],axis=1)
Y=df['output-PV Yield (kWh)']
pca = PCA(n_components=9)
pca.fit(X_train)
X_train = pca.transform(X_train)
pca.fit(X_test)
X_test = pca.transform(X_test)
#normalizing the input values to fall in -1 to 1
X_train = X_train/180000000.0
X_test = X_test/180000000.0
#Creating Model
model = Sequential()
model.add(Dense(15, input_shape=(9,)))
model.add(Activation('tanh'))
model.add(Dense(11))
model.add(Activation('tanh'))
model.add(Dense(1))
model.summary()
sgd = optimizers.SGD(lr=0.1,momentum=0.2)
model.compile(loss='mean_absolute_error',optimizer=sgd,metrics=['accuracy'])
#Training
model.fit(X_train, train_y, epochs=20, batch_size = 50, validation_data=(X_test, test_y))
My weights are not getting updated. Accuracy is zero in all epochs.
The model seems OK but there are two problems I can spot fast:
pca = PCA(n_components=9)
pca.fit(X_train)
X_train = pca.transform(X_train)
pca.fit(X_test)
X_test = pca.transform(X_test)
Anything used for transformation of the data must not be fit on testing data. You fit it on train samples and then use it to transform both train and test part. You should assume that you know nothing about data you will be predicting on in production, eg. you know nothing about tomorrows weather, results of sport matches in a month, etc. You wont be able to do so then, so you cant do so during training. Correct way:
pca = PCA(n_components=9)
pca.fit(X_train)
X_train = pca.transform(X_train)
X_test = pca.transform(X_test)
The second very incorrect stuff you have there is here:
#normalizing the input values to fall in -1 to 1
X_train = X_train/180000000.0
X_test = X_test/180000000.0
Of course you want to normalize your data, but this way you will end up with incredibly low decimals in cases where values are low, eg. AlbedoDaily column, and quite high values where are values high, such as SurfacePressure. For such scaling you can use already defined classes such as standard scaler. The code is very simple and each column is treated independently:
from sklearn.preprocessing import StandardScaler
transformer = StandardScaler().fit(X_train)
X_train = transformer.transform(X_train)
X_test = transformer.transform(X_test)
You have not provided or explained what your target variable is and where you get is, there could be other problems in your code I can not see right now.

Predicting next value of game results

I am working on predicting game output. The game has three seats. Seat A , seat B and seat C. Each time only one seat will win randomly.
I have written LSTM code to predict the next value. I made the dataset of around 8000 games and recorded the results. But LSTM is not predicting the next value. It is either predicting either all "0"s or all "1"s. I'm curious why it is not learning.
here you can find dataset
here is attached my google colab code
# univariate data preparation
from sklearn.preprocessing import OneHotEncoder
from numpy import array
import pandas as pd
# split a univariate sequence into samples
df = pd.read_csv('DatasetLarge.csv') # here you can find dataset https://drive.google.com/open?id=1WZBMYO-Oi3uErPlBXphXFAACmn9VK_yZ
def split_sequence(sequence, n_steps):
X, y = list(), list()
for i in range(len(sequence)):
# find the end of this pattern
end_ix = i + n_steps
# check if we are beyond the sequence
if end_ix > len(sequence)-1:
break
# gather input and output parts of the pattern
seq_x, seq_y = sequence[i:end_ix], sequence[end_ix]
X.append(seq_x)
y.append(seq_y)
return array(X), array(y)
# Removing some errors of dataset
df=df.replace(to_replace ="q",
value ="a")
df=df.replace(to_replace ="aa",
value ="a")
df=df.replace(to_replace ="cc",
value ="c")
#label encoding
from sklearn import preprocessing
le = preprocessing.LabelEncoder()
le.fit(df.values)
df=le.transform(df)
# define input sequence
raw_seq=df
# choose a number of time steps
n_steps = 5
# split into samples
X, y = split_sequence(raw_seq, n_steps)
# summarize the data
for i in range(len(X)):
print(X[i], y[i])
#Spliting dataset for train and test
X_train=X[0:7500]
y_train=y[0:7500]
X_test=X[7500:]
y_test=y[7500:]
import tensorflow.keras
from tensorflow.keras.models import Sequential, Model
from tensorflow.keras.layers import Dense, Dropout, Flatten,Input,LSTM
n_features = 1
X_train = X_train.reshape((X_train.shape[0], X_train.shape[1], n_features))
# define model
model = Sequential()
model.add(LSTM(200, activation='softmax', input_shape=(n_steps, n_features)))
model.add(Dense(1))
model.compile(optimizer='adam', loss='binary_crossentropy')
# # fit model
model.fit(X_train, y_train, epochs=100, verbose=0)
# # demonstrate prediction
X_test=X_test.reshape((X_test.shape[0], X_test.shape[1], n_features))
yhat_classes =(model.predict(X_test) > 0.5).astype("int32")
print(yhat_classes)
Please help me why it is not predicting correctly.

Overfitting and data leakage in tensorflow/keras neural network

Good morning, I'm new in machine learning and neural networks. I am trying to build a fully connected neural network to solve a regression problem. The dataset is composed by 18 features and 1 label, and all of these are physical quantities.
You can find the code below. I upload the figure of the loss function evolution along the epochs (you can find it below). I am not sure if there is overfitting. Someone can explain me why there is or not overfitting?
import pandas as pd
import numpy as np
from sklearn.ensemble import RandomForestRegressor
from sklearn.feature_selection import SelectFromModel
from sklearn import preprocessing
from sklearn.model_selection import train_test_split
from matplotlib import pyplot as plt
import keras
import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Dropout, BatchNormalization
from tensorflow.keras.callbacks import EarlyStopping
from keras import optimizers
from sklearn.metrics import r2_score
from keras import regularizers
from keras import backend
from tensorflow.keras import regularizers
from keras.regularizers import l2
# =============================================================================
# Scelgo il test size
# =============================================================================
test_size = 0.2
dataset = pd.read_csv('DataSet.csv', decimal=',', delimiter = ";")
label = dataset.iloc[:,-1]
features = dataset.drop(columns = ['Label'])
y_max_pre_normalize = max(label)
y_min_pre_normalize = min(label)
def denormalize(y):
final_value = y*(y_max_pre_normalize-y_min_pre_normalize)+y_min_pre_normalize
return final_value
# =============================================================================
# Split
# =============================================================================
X_train1, X_test1, y_train1, y_test1 = train_test_split(features, label, test_size = test_size, shuffle = True)
y_test2 = y_test1.to_frame()
y_train2 = y_train1.to_frame()
# =============================================================================
# Normalizzo
# =============================================================================
scaler1 = preprocessing.MinMaxScaler()
scaler2 = preprocessing.MinMaxScaler()
X_train = scaler1.fit_transform(X_train1)
X_test = scaler2.fit_transform(X_test1)
scaler3 = preprocessing.MinMaxScaler()
scaler4 = preprocessing.MinMaxScaler()
y_train = scaler3.fit_transform(y_train2)
y_test = scaler4.fit_transform(y_test2)
# =============================================================================
# Creo la rete
# =============================================================================
optimizer = tf.keras.optimizers.Adam(lr=0.001)
model = Sequential()
model.add(Dense(60, input_shape = (X_train.shape[1],), activation = 'relu',kernel_initializer='glorot_uniform'))
model.add(Dropout(0.2))
model.add(Dense(60, activation = 'relu',kernel_initializer='glorot_uniform'))
model.add(Dropout(0.2))
model.add(Dense(60, activation = 'relu',kernel_initializer='glorot_uniform'))
model.add(Dense(1,activation = 'linear',kernel_initializer='glorot_uniform'))
model.compile(loss = 'mse', optimizer = optimizer, metrics = ['mse'])
history = model.fit(X_train, y_train, epochs = 100,
validation_split = 0.1, shuffle=True, batch_size=250
)
history_dict = history.history
loss_values = history_dict['loss']
val_loss_values = history_dict['val_loss']
y_train_pred = model.predict(X_train)
y_test_pred = model.predict(X_test)
y_train_pred = denormalize(y_train_pred)
y_test_pred = denormalize(y_test_pred)
plt.figure()
plt.plot((y_test1),(y_test_pred),'.', color='darkviolet', alpha=1, marker='o', markersize = 2, markeredgecolor = 'black', markeredgewidth = 0.1)
plt.plot((np.array((-0.1,7))),(np.array((-0.1,7))),'-', color='magenta')
plt.xlabel('True')
plt.ylabel('Predicted')
plt.title('Test')
plt.figure()
plt.plot((y_train1),(y_train_pred),'.', color='darkviolet', alpha=1, marker='o', markersize = 2, markeredgecolor = 'black', markeredgewidth = 0.1)
plt.plot((np.array((-0.1,7))),(np.array((-0.1,7))),'-', color='magenta')
plt.xlabel('True')
plt.ylabel('Predicted')
plt.title('Train')
plt.figure()
plt.plot(loss_values,'b',label = 'training loss')
plt.plot(val_loss_values,'r',label = 'val training loss')
plt.xlabel('Epochs')
plt.ylabel('Loss Function')
plt.legend()
print("\n\nThe R2 score on the test set is:\t{:0.3f}".format(r2_score(y_test_pred, y_test1)))
print("The R2 score on the train set is:\t{:0.3f}".format(r2_score(y_train_pred, y_train1)))
from sklearn import metrics
# Measure MSE error.
score = metrics.mean_squared_error(y_test_pred,y_test1)
print("\n\nFinal score test (MSE): %0.4f" %(score))
score1 = metrics.mean_squared_error(y_train_pred,y_train1)
print("Final score train (MSE): %0.4f" %(score1))
score2 = np.sqrt(metrics.mean_squared_error(y_test_pred,y_test1))
print(f"Final score test (RMSE): %0.4f" %(score2))
score3 = np.sqrt(metrics.mean_squared_error(y_train_pred,y_train1))
print(f"Final score train (RMSE): %0.4f" %(score3))
EDIT:
I tried alse to do feature importances and to raise n_epochs, these are the results:
Feature Importance:
No Feature Importace:
Looks like you don't have overfitting! Your training and validation curves are descending together and converging. The clearest sign you could get of overfitting would be a deviation between these two curves, something like this:
Since your two curves are descending and are not diverging, it indicates your NN training is healthy.
HOWEVER! Your validation curve is suspiciously below the training curve. This hints a possible data leakage (train and test data have been mixed somehow). More info on a nice an short blog post. In general, you should split the data before any other preprocessing (normalizing, augmentation, shuffling, etc...).
Other causes for this could be some type of regularization (dropout, BN, etc..) that is active while computing the training accuracy and it's deactivated when computing the Validation/Test accuracy.
Overfitting is, when the model does not generalize to other data than the training data. When this happen you will have a very (!) low training loss but a high validation loss. You can think of it this way: if you have N points you can fit a N-1 polynomial such that you have a zero training loss (your model hits all your training points perfectly). But, if you apply that model to some other data, it will most likely produce a very high error (see the image below). Here the red line is our model and the green is the true data (+ noice), and you can see in the last picture we get zero training error. In the first, our model is too simple (high train/high validation error), the second is good (low train/low valuidation error) the third and last is too complex i.e overfitting (very low train/high validation error).
Neural network can work in the same way, so by looking at your training vs validation error, you can conclude if it overfits or not
No, this is not overfitting as your validation loss isn´t increasing.
Nevertheless, if I were you I would be a little bit skeptical. Try to train your model for even more epochs and watch out for the validation loss.
What you definitely should do, is to observe the following:
- are there duplicates or near-duplicates in the data (creates information leakage from train to test validation split)
- are there features that have a causal connection to the target variable
Edit:
Usually, you have some random component in a real-world dataset, so that rules that are observed in train data aren´t 100% true for validation data.
Your plot shows that the validation loss is even more decreasing as train loss decreases. Usually, you get to some point in training, where the rules you observe in train data are too specific to describe the whole data. That´s when overfitting begins. Hence, it is weird, that your validation loss doesn´t increase again.
Please check whether your validation loss approaches zero when you´re training for more epochs. If it´s the case I would check your database very carefully.
Let´s assume, that there is a kind of information leakage from the train set to the validation set (through duplicate records for example). Your model would change the weights to describe very specific rules. When applying your model to new data it would fail miserably since the observed connections are not really general.
Another common data problem is, that features may have an inversed causality.
The thing that validation loss is generally lower than train error is probably depending on dropout and regularization, since it´s applied while training but not for predicting/testing.
I put some emphasis on this because a tiny bug or an error in the data can "fuck up" your whole model.

Categories